Does the METHOD study show that ZOE works? (2024)

“The evidence is in: ZOE works.” The email from ZOE, sent to people who had signed up for information about it, went on “Most health and wellness companies don’t test their products scientifically. So, there’s no proof that they work. This is where we’re different. As a science-first company, we decided to put ZOE to the test with a rigorously conducted, randomised controlled trial.” The email from ZOE to potential customers was timed to the publication of their much trailed research paper, the METHOD study (Measuring Efficacy Through Outcomes of Diet), in Nature Medicine.1

But does the METHOD study show that ZOE “works”? The tech company, co-founded by Tim Spector, professor at King’s College London, and promoted by Dragons Den star Steve Barlett, as well as other celebrities, offers customers—for a fee of around £299 to £599 for 12 months—a “personalised” diet programme after analysing blood samples, a stool test for microbiome, and a period of continuous blood glucose monitoring.2 Based on these, customers get recommendations of what to eat—and what not to. The email from ZOE told customers that “Compared with the control group, ZOE members saw greater improvements in a range of health measures. For instance, they had increases in ‘good’ gut bacteria linked to better health outcomes. They also reported feeling less hungry, sleeping better, and having more energy.”

Is this what the study showed? To prove that the personalised dietary advice, based on the continuous glucose monitoring and microbiome assays, had an impact, the only difference between comparative groups should be that—dietary advice based on these results. But this isn’t what the study did. The study ran over 18 weeks in US adults. The control group got standard US diet advice, via a leaflet and an online video. They also received a weekly generic email.

The intervention group got something quite different. They had interactive “lessons” on the app, explaining concepts such as fibre, wholegrains, and diverse diets. They had a microbiome assay, and test foods while using continuous glucose monitoring, which resulted in “personalised information” via their smart phones, and a personalised results walk-through with a member of the research team, specific meal recommendations, programme lessons, and repeated completion of food diaries. It would have been obvious to the control group who they were. The differences between the two interventions are so distinct that it is impossible to say whether ZOE’s “USP,” the “personalised advice” generated by the stool and glucose test, had a bearing on the results. If the control group had, say, a continuous glucose monitor, but with the results blinded to the researchers and the participants, who were then given apparently personalised, but random recommendations for particular foods, with all the same support and coaching, it would have been a better comparison.

In studies of dietary interventions, tailored support can have large impacts on outcomes. Trial design is crucial because of the potential for Hawthorne effects (the impact of measurement on outcomes), expectation bias, and placebo effects.3 The ZOE brand is now well established and widely publicised—it would have been difficult for participants to be unaware of it. Even hunger ratings can be affected by placebo effects.4 The primary outcomes of the study were triglyceride and low-density lipoprotein cholesterol (LDL-C) concentration; secondary outcomes were weight, waist and hip size, blood pressure, HbA1c, multiple other blood results, and self reported mood, hunger, and “energy level.” The control group gained a little weight by the end of the trial, with a difference of 2.46 kg compared with the intervention group, who also lost 2.35 cm on average from their waist circumference. There was no change in one primary outcome measure—LDL-C concentration. But there was a statistically significant change in triglycerides (a decrease of 0.13 mmol/l), although the clinical significance of this is questionable.

High quality clinical trials in diet are difficult to do. Attempts to improve the research base in this area are to be applauded. However, this is undermined when claims go beyond what is demonstrated, as is the case here. Further, the trial was funded by ZOE, and each of the authors is linked to ZOE as a co-founder, consultant, employee, option holder, or member of ZOE’s scientific board. This was, therefore, not an independent study. Claims about benefit should remain guarded while the evidence base remains so poor.